World's most intelligent
World's most intelligent
retrieval layer for AI
World's most
intelligent
intuitive
complete
powerful
intelligent
intelligent
intuitive
complete
powerful
intelligent
retrieval layer for AI
retrieval layer for AI
Cortex is a self-improving search engine that eliminates fine-tuning vector databases, selecting embeddings, configuring knowledge graphs, or building custom ingestion pipelines for different data.
The perfect search that makes AI work flawlessly.
Cortex is a self-improving search engine that eliminates efforts wasted on fine-tuning vector databases, selecting embeddings, configuring knowledge graphs, or building custom ingestion pipelines for different data.
The perfect search for AI.
Cortex delivers a self-improving search engine that eliminates efforts wasted on
fine-tuning vector databases, selecting embeddings, configuring knowledge graphs, setting up memory, or building custom ingestion pipelines for data.
The perfect search for AI, without the chaos.
Proudly powering AI companies at the frontier
Powering frontier AI at
COMPARISON
Build intelligent AI
How developers build successful AI products
Self-improving Search (vector + full text + keyword)
Intelligent metadata querying before search to ensure complete coverage and extremely high accuracy
Ingestion engine with native parsing, chunking, and processing built-in
Low-latency search even at scale (p50 < 50ms, p95 < 200ms)
Built-in memory that self-improves with usage
Embedding pipelines that auto scale to petabytes of data
AI-generated answers with 20+ configurations (like stream, language models, recency_bias, multi_step_reasoning)
Traditional VectorDBs
Do not improve with every user interaction
Always search across entire corpus leading to
incomplete results
Bring-your-own-parsers mess—manually monitor every change in structure or format
High latencies—often >1s at p50 under real-world workloads
No memory or personalization
Manual, brittle embedding pipelines that break at scale
No support or control over generation behavior, reasoning steps, or context injection
Goodbye complexity
Vector search doesn't work. Your agents need an adaptive retrieval layer that provides personalization with every user query, delivers accurate results, and makes your AI app memorable.
Easiest to deploy
Deploy in minutes, not months
Extremely powerful search, serverless ingestion that works at scale, built-in memory for agents, self improving responses. New paradigm of search for AI.
from usecortex_ai import CortexAI, AsyncCortexAI # supports both sync and async clients
# Initialize client
client = CortexAI(token=api_key)
# Create a tenant
client.user.create_tenant(tenant_id=tenant_id)
# Upload a document
client.upload.upload_document(
tenant_id=tenant_id,
file=some_file
)
# Upload plain text
client.upload.upload_text(
tenant_id=tenant_id,
request=MarkdownUploadRequest(
content="This is a sample text document for testing purposes. "
"It contains information about artificial intelligence and machine learning."
),
)
# Retrieve search results
client.search.retrieve(
tenant_id=tenant_id,
query="What is cortex-ai",
max_chunks=10
)
# Full-text search
client.search.full_text_search(
tenant_id=tenant_id,
query="artificial intelligence",
max_chunks=10
)
# Get tenant stats
client.tenant.stats(tenant_id=tenant_id)
# Verify file processing
client.upload.verify_processing(
tenant_id=tenant_id,
file_id="CortexDoc8c396f0198304713b3fb340c6075ce571756537873"
)from cortex import CortexClient
cortex = CortexClient(api_key="your-key")
# 1. SEARCH: Search across all sources
results = cortex.search("latest partnership details with Umbrella Corp")
# → ["Found recent documents discussing the Umbrella Corp deal...", ...]
# 2. QnA: Ask contextual questions
answer = cortex.answer("What are the terms of the partnership?")
# → "The Umbrella Corp partnership includes a joint GTM plan and 12-month exclusivity in biotech integrations."
# 3. STORE: Add new knowledge to the system
cortex.store(
user_id="tenant1",
document_title="Umbrella Partnership Summary",
content="Joint GTM strategy, 12-month exclusivity, and integration support included."
)
# 4. REMEMBER: Store personalized memory
cortex.remember(
user_id="tenant1",
content="I’m meeting Umbrella Corp next Thursday to finalize integration timelines."
)
# LATER: Ask anything and get memory-aware answers
response = cortex.answer("Remind me what I need to finalize in the Umbrella meeting?")
# → "You're meeting Umbrella Corp next Thursday to finalize integration timelines."
# → "They expect joint GTM rollout by Q4. You’ve stored details about exclusivity and support."Fully Stacked
Everything you need to instantly make your AI memorable
Hybrid search for AI
Intelligent vector, full-text, and metadata search. Complete with optimal retrieval and reranking to return the right context. Audit logs explain every result.
Hybrid search for AI
Intelligent vector, full-text, and metadata search. Complete with optimal retrieval and reranking to return the right context. Audit logs explain every result.
Hybrid search for AI
Intelligent vector, full-text, and metadata search. Complete with optimal retrieval and reranking to return the right context. Audit logs explain every result.
Agentic metadata search
Filter by any field from your systems, not just what is inside the index. Build exact lists, see every document in a category, and get the precise details for each document.
Agentic metadata search
Filter by any field from your systems, not just what is inside the index. Build exact lists, see every document in a category, and get the precise details for each document.
Agentic metadata search
Filter by any field from your systems, not just what is inside the index. Build exact lists, see every document in a category, and get the precise details for each document.
Smart ingestion and segmentation
Connect Slack, Gmail, Notion, PDFs, tickets, and more. We parse, clean, and segment content per source so chunks match how people read.
Smart ingestion and segmentation
Connect Slack, Gmail, Notion, PDFs, tickets, and more. We parse, clean, and segment content per source so chunks match how people read.
Smart ingestion and segmentation
Connect Slack, Gmail, Notion, PDFs, tickets, and more. We parse, clean, and segment content per source so chunks match how people read.
Built-in memory and personalization
Results improve with every query. Cortex learns from interactions per user and per tenant to boost relevance over time.
Built-in memory and personalization
Results improve with every query. Cortex learns from interactions per user and per tenant to boost relevance over time.
Built-in memory and personalization
Results improve with every query. Cortex learns from interactions per user and per tenant to boost relevance over time.
Complete SDK for AI
The first SDK and APIs for search, retrieval, and answers. Add assistants, tool calls, and structured responses in minutes. Trace queries, understand ranking, and enforce permissions.
Complete SDK for AI
The first SDK and APIs for search, retrieval, and answers. Add assistants, tool calls, and structured responses in minutes. Trace queries, understand ranking, and enforce permissions.
Complete SDK for AI
The first SDK and APIs for search, retrieval, and answers. Add assistants, tool calls, and structured responses in minutes. Trace queries, understand ranking, and enforce permissions.
Multimodal retrieval
Run high-throughput embedding pipelines that scale to massive datasets without breaking.
Multimodal retrieval
Run high-throughput embedding pipelines that scale to massive datasets without breaking.
Multimodal retrieval
Run high-throughput embedding pipelines that scale to massive datasets without breaking.
Security
Enterprise Grade Compliance
SOC 2 compliant, self-hostable, and built for enterprise. Stay in control of your data.
Cortex is built with privacy at its core. As a SOC 2 certified platform, our entire architecture and codebase can be audited at any time, making us one of the most transparent and secure options available, almost like open-sourcing our security.
PRICING
Flexible Plans for Every AI
Flexible Plans for Every AI
From idea to enterprise. Cortex scales with you.
From idea to enterprise. Cortex scales with you.
Monthly
Yearly
Save 20%
Starter
$450
/month
Self improving search (hybrid, vector, full text)
10M tokens of ingestion + storage / month
Upto 10 Databases
50,000 monthly active users
Slack Support
Pro
$5000
/month
100M tokens of ingestion + storage / month
Unlimited Databases & MAUs
SOC 2
Early access to new features and models
Priority Support
Enterprise
Custom
Everything in Pro, plus:
Dedicated AI Strategist
API & Private Integrations
On Premises deployment
Custom SLA
Complete control of your data
Monthly
Yearly
Save 20%
Starter
$450
/month
Self improving search (hybrid, vector, full text)
10M tokens of ingestion + storage / month
Upto 10 Databases
50,000 monthly active users
Slack Support
Pro
$5000
/month
100M tokens of ingestion + storage / month
Unlimited Databases & MAUs
SOC 2
Early access to new features and models
Priority Support
Enterprise
Custom
Everything in Pro, plus:
Dedicated AI Strategist
API & Private Integrations
On Premises deployment
Custom SLA
Complete control of your data
FAQ'S
Frequently Asked Questions
Still Have Questions?
Get in touch - hello@usecortex.ai
What is Cortex and how is it different from a traditional vector database?
Cortex is more than a vector DB — it’s the complete retrieval layer for AI-native apps. It handles ingestion, parsing, chunking, embedding, hybrid search (vector + keyword + metadata), memory, and grounding — all out-of-the-box. No need to stitch together 20 tools to get reliable retrieval.
Can I bring any data and connect multiple sources?
Yes. Cortex supports ingestion from files (PDFs, HTML, etc.), APIs, databases, Notion, Slack, and more. Just connect your source — Cortex automatically applies the best parsing and embedding strategy for your data type, then handles chunking, syncing, and storage end-to-end.
How customizable is the AI-generated output?
Highly customizable. Our API supports 20+ parameters like search_modes, recency_bias, multi_step_reasoning, highlight_chunks, user_instructions, and more — giving you full control over how answers are retrieved and generated.
What does performance look like at scale?
Cortex is built for low-latency, high-scale workloads. You get <50ms p50 and <200ms p95 response times, even with large datasets and multi-modal retrieval. No babysitting infrastructure required.
Why does Cortex support memory and personalization?
Because great AI isn’t one-size-fits-all. Cortex builds memory into the retrieval layer so your agents and apps can adapt to users over time — remembering past interactions, surfacing relevant context, and tailoring responses based on behavior or metadata. Personalization isn’t an add-on — it’s core to making AI actually useful.
FAQ'S
Frequently Asked Questions
Still Have Questions?
Get in touch - hello@usecortex.ai
What is Cortex and how is it different from a traditional vector database?
Cortex is more than a vector DB — it’s the complete retrieval layer for AI-native apps. It handles ingestion, parsing, chunking, embedding, hybrid search (vector + keyword + metadata), memory, and grounding — all out-of-the-box. No need to stitch together 20 tools to get reliable retrieval.
Can I bring any data and connect multiple sources?
Yes. Cortex supports ingestion from files (PDFs, HTML, etc.), APIs, databases, Notion, Slack, and more. Just connect your source — Cortex automatically applies the best parsing and embedding strategy for your data type, then handles chunking, syncing, and storage end-to-end.
How customizable is the AI-generated output?
Highly customizable. Our API supports 20+ parameters like search_modes, recency_bias, multi_step_reasoning, highlight_chunks, user_instructions, and more — giving you full control over how answers are retrieved and generated.
What does performance look like at scale?
Cortex is built for low-latency, high-scale workloads. You get <50ms p50 and <200ms p95 response times, even with large datasets and multi-modal retrieval. No babysitting infrastructure required.
Why does Cortex support memory and personalization?
Because great AI isn’t one-size-fits-all. Cortex builds memory into the retrieval layer so your agents and apps can adapt to users over time — remembering past interactions, surfacing relevant context, and tailoring responses based on behavior or metadata. Personalization isn’t an add-on — it’s core to making AI actually useful.
FAQ'S
Frequently Asked Questions
Still Have Questions?
Get in touch - hello@usecortex.ai
What is Cortex and how is it different from a traditional vector database?
Can I bring any data and connect multiple sources?
How customizable is the AI-generated output?
What does performance look like at scale?
Why does Cortex support memory and personalization?